During vertebrate evolution, there has been a distinct switch from a strong hormonal influence on sex determination to a genetic switch mechanism. In nonmammalian vertebrates, oestrogen administration to male embryos, or blocking oestrogen production in female embryos, results in sex reversal [1]. In contrast, in mice, oestrogen does not affect the early development of either ovaries or testes, and in its absence, gonadal development proceeds normally until after birth [2]. Although the role of oestrogen in determining gonadal fate differs between vertebrates, the downstream pathways of gonadal differentiation are highly conserved. In particular, SOX9, the key regulator of testicular differentiation, appears to have similar functions in all vertebrates [3], starting with cytoplasmic expression in the indifferent gonad, but translocating to the nucleus in males to activate the testicular pathway [4–6]. In oestrogen-deficient female mice (made by deletion of one or both oestrogen receptor genes or of the Cyp19A1 gene that encodes the aromatase enzyme), germ cells are gradually lost from pubertal ovaries concomitant with remodeling of the somatic tissue. There is also transdifferentiation of the ovarian granulosa cells to form testicular Sertoli-like cells that express SOX9 and align in structures resembling testicular cords [7]. This phenotype can be reversed by administration of oestrogen, causing the somatic cells to reform granulosa cells and resulting in an ovarian architecture and suppression of SOX9 [7]. These data show that although oestrogen is not required for early ovarian development in mice, it is critical for maintaining ovarian somatic cell fate in the differentiated gonad.

How and why oestrogen became dissociated from sex determination in mammals is unknown. A genetically controlled sex determination mechanism may have evolved in mammals to protect the young developing in utero from the effect of large quantities of oestrogen in the fetoplacental unit [8, 9]. However, oestrogen receptor expression is maintained in the indifferent mammalian gonad [10, 11] despite their apparent lack of function, and exposure of gonads to exogenous oestrogen activates the translocation of the receptors to the nuclei of indifferent gonadal cells [12]. Not surprisingly, the mammalian placenta is highly effective in metabolizing hormones before they reach the developing fetus [13], protecting it from maternal influences. Marsupial mammals are born with undifferentiated gonads that undergo gonadal sex determination postnatally [14], allowing daily administration of oestrogen directly to developing male young from before the testis pathway is initiated. Oestrogen treatment of XY young in the opossum and in the tammar wallaby directs the development of the indifferent gonads towards an ovarian differentiation fate [15–21], including entry of the XY germ cells into meiosis, despite the presence of a Y-chromosome and SRY [17–19, 21].

The sex-reversing effects of exogenous oestrogen on the marsupial XY gonad are time-dependent [17, 18], just as there is a critical time window of Sry action in the mouse [22]. Complete ovarian formation with germ cells arrested in meiosis was seen only in animals that were born about 1 day prematurely in a 26.5-day pregnancy in the tammar [18, 21] and one half to 1 day prematurely in the 13.5-day gestation in the opossum [17]. Treatment of animals born after a full-term pregnancy with the same dose of oestrogen did not affect testicular development and resulted in only minor disruption of the testis in both species [17–19].

These findings suggest that there is a critical point in development when the XY gonad becomes irreversibly committed to a testicular fate. In the tammar, this coincides with the first morphological signs of Sertoli cell development and seminiferous cord formation, which occurs after birth from a normal full-term pregnancy [14]. To promote ovarian development, oestrogen must be administered before the somatic cell lineage has initiated Sertoli cell development. Early experiments treating opossums with large amounts of androgens did not affect gonadal development of male or female young [15, 16]. Similarly, in the tammar, testosterone administration to male pouch young had no visible effect on testicular structure, although oestrogen given after testes development was initiated did cause a small degree of testicular dysgenesis [20]. Identical findings have also been reported in nonmammalian vertebrates in which administration of oestrogen to the indifferent gonad can actively induce ovarian development, independent of genetic or environmental sex-determining factors [1, 23], whereas exogenous androgen treatment does not affect testicular development [23]. The effects of oestrogen are mediated through the oestrogen receptors (ERs) in the gonad [2]. Treatment of the tammar XY gonads at day 25 of gestation with oestrogen in culture causes the oestrogen receptors to move from a cytoplasmic to nuclear localization within the cell, consistent with their activation and ability to affect gene transcription in an oestrogen-dependent manner [12]. The nuclear translocation of both receptor proteins in the presence of exogenous oestrogen in culture confirms the ability of the receptors to respond functionally to the ligand, even at early developmental stages [12]. Thus the observed effects of exogenous oestrogen are oestrogen-, and not steroid-, specific.

The genes controlling sex determination are conserved between eutherian and marsupial mammals. SRY and SOX9 are expressed in the developing Sertoli cells of the XY eutherian and marsupial testis [4, 24, 25]. These cells direct the development of the testicular architecture and secrete AMH, a hormone necessary for the regression of the Müllerian ducts [26–28]. In the absence of SRY in developing XX gonads, SOX9 levels are significantly lower [4, 25], AMH is not activated and the Müllerian ducts remain intact [26, 27]. FGF9 may also be important for Sertoli cell specification [29] and shares an antagonistic relationship with an ovarian differentiation gene, WNT4 [30]. Other key ovarian development genes include FOXL2, RSPO1 and FST, which are upregulated in the somatic cells of the early developing ovary in mice and humans [31–33]. Disruption of each gene dramatically affects ovarian differentiation and can lead to varying degrees of testicular development in XX mammals [31–33].

To determine the molecular changes occurring during oestrogen -induced sex reversal of XY gonads, we investigated the expression of key male and female differentiation factors (SRY, SOX9, AMH and FGF9 and key ovarian markers FOXL2, FST, RSPO1 and WNT4) in our model marsupial, the tammar wallaby, during the critical period of gonadal differentiation and after oestrogen-induced sex reversal. Here we show that exposure of the indifferent XY gonad to oestrogen resulted in suppression of SRY and AMH and upregulation of FOXL2 and WNT4, thus diverting the male developmental pathway into the female one. The SOX9 gene was not downregulated by oestrogen exposure during XY ovarian development, but SOX9 protein was prevented from entering the nucleus of the supporting cells in the presence of oestrogen and therefore could not activate the male developmental program. We hypothesise that this novel function of oestrogen in modulating the nucleocytoplasmic localization of SOX9 may reflect an ancestral primordial mechanism by which oestrogen mediates sex determination in nonmammalian development. However, oestrogen retains an important function in adult female mammals by maintaining granulosa cell fate in the differentiated ovary by suppressing nuclear localization of the SOX9 protein.

Marsupials occupy a unique evolutionary niche and can provide important information on the evolution and conservation of mammalian developmental pathways. This is the first comprehensive study of the expression levels of key testicular and ovarian genes during gonadogenesis in any marsupial. During normal development in the tammar gonad, expression of key male development genes is similar to that seen in the mouse and human, except that AMH and SOX9 show sexually dimorphic gene expression before the peak in SRY expression at day 1 pp. Furthermore, SOX9 has already begun its nuclear translocation 3 days before the peak in SRY expression and the appearance of testicular cords. This may be the result of the long period of SRY expression seen in marsupials [24], similar to that of human XY gonads [36] but different from the ultrashort pulse seen in mice [34, 35]. The key female development genes FOXL2 and WNT4 were upregulated in the XX gonad during the time of ovarian development (5-10 days pp). However, RSPO1 and FST expression was variable in the early gonad, and female-specific upregulation was seen only during the later stages of ovarian development. This observation suggests that these two genes act downstream of FOXL2 and WNT4 in the ovarian differentiation pathway, at least in the tammar.

Cultured XY control gonads had normal male expression profiles and formed testicular cords. However, XY gonads cultured in the presence of oestrogen failed to upregulate SRY and AMH expression and instead upregulated FOXL2 and WNT4. The oestrogen-exposed XY gonads also developed ovarian architecture, confirming that female development had occurred. Surprisingly, mRNA levels for SOX9, the key factor essential for testicular development, did not differ between control and treated XY gonads, despite ovarian development. To investigate this further, we examined the protein localisation of SOX9 in control and oestrogen-treated XY gonads. Since SOX9 is a transcription factor, nuclear localisation is essential to activate the male developmental program and SOX9 mutations that affect nuclear import result in XY sex reversal in humans [41, 42]. In the mouse and human, the SOX9 protein is initially cytoplasmic in the indifferent gonad of both XX and XY fetuses, but becomes rapidly nuclear and upregulated only after the peak of SRY expression in testes, whereas in ovaries, Sox9 remains cytoplasmic and is lost [4, 5]. In the clawed toad, Xenopus, SOX9 also remains cytoplasmic in developing ovaries but becomes nuclear under testis-promoting conditions in the absence of oestrogen and is upregulated [6, 43]. However, in the presence of oestrogen under ovary-promoting conditions, SOX9 remains cytoplasmic and is downregulated.

In mice, Sry initiates Sox9 upregulation within pre-Sertoli cells. Sox9 then autoregulates to reach threshold levels that activate testicular development [44]. In the tammar, SRY had initiated SOX9 upregulation before the cultures began (Fig. 1; Additional file 1) and before the peak in SRY expression 3 days later. Thus oestrogen exposure at this time (day 25 of gestation) did not prevent the initial wave of SOX9 mRNA upregulation, but the protein was restricted to the cytoplasm of somatic cells, preventing its activation of the male pathway and likely its self-propagated upregulation. This observation explains the initiation of ovarian development in the treated XY gonads in this in vitro study and in the earlier in vivo studies [17, 18, 21]. Treatment of developing turtle gonads with oestrogen at stages of development before gonadal sex determination had occurred (and prior to the initial upregulation of SOX9) prevented SOX9 from becoming upregulated to equivalent male or even female levels [45]. These findings suggest that there is also a direct action of oestrogen in repressing SOX9 transcription in turtles, as we have observed in the tammar. Interestingly, in the turtle, SOX9 protein is nuclear in female gonads before sex determination. We see a similar nuclear localization in the indifferent gonad in the tammar wallaby (Additional file 3). We hypothesize that in the absence of hormones at early stages of gonadal development, SOX9 can feely translocate to the nucleus of somatic cells. In mammals, SOX9 expression is reinforced by SRY, upregulating it above threshold levels enabling it to autoregulate and initiate testicular development. In nonmammalian vertebrates, the oestrogen production could act to directly suppress SOX9 upregulation and promote ovarian development. Our results show that oestrogen restricts translocation of the SOX9 protein, forcing it to remain in the cytoplasm and prevent its self-propagated upregulation. Since SOX9 mRNA is never elevated above basal levels in the oestrogen-treated male turtle, the effect of oestrogen on its subcellular localization cannot be observed in their gonads. However, in our study, we administered oestrogen slightly later in development, just after the initial wave of SOX9 upregulation. Our results therefore highlight a mechanism by which oestrogen could prevent testicular development (Fig. 4). The exclusion of SOX9 from somatic cell nuclei of the oestrogen cultured gonads is further confirmed by the failure of AMH upregulation because in the mouse, Sox9 acts as a nuclear transcription factor to drive Amh expression [46].

In the absence of nuclear SOX9 in XY oestrogen-treated gonads, female development was initiated. FOXL2 was significantly upregulated, consistent with an early and essential role for this gene in initiating ovarian development in mammals. In goats, FOXL2 induces aromatase (CYP19) expression, leading to oestrogen production in the developing ovary [38]. This mechanism could potentially ensure a block in male development by excluding SOX9 nuclear entry. This mechanism could work in synergy with activated ERs to facilitate FOXL2 binding to TESCO, further suppressing SOX9 mRNA production as recently suggested in goats [38]. However, the oestrogen response elements contained within the mouse TESCO element [37] do not fall into the evolutionarily conserved regions identified in most vertebrates [3], suggesting that this mechanism of oestrogen action may not be as important as its role to prevent SOX9 from entering the nucleus.

Several lines of evidence demonstrate that oestrogen acts as a modulator of the subcellular localization of SOX9. Sox9 is produced at a low level in the somatic cells of the adult mouse ovary [47], but its self-propagated upregulation is somehow blocked. Our findings suggest that circulating oestrogen could prevent the nuclear translocation of SOX9 in differentiated ovarian somatic cells, thus preventing its own upregulation. In the absence of oestrogen in female aromatase-deficient knockout mice, Sox9 is upregulated in ovarian somatic cells, as it can presumably translocate to the nucleus to propagate its own transcription. However, when exogenous oestrogen is administered to aromatase-deficient mice, this effect is reversible and Sox9 is repressed [5, 48]. This mechanism therefore explains how Sox9 upregulation is able to occur in the oestrogen-deficient mouse ovary and how it can be subsequently suppressed with somatic cell transdifferentiation in the presence of exogenous oestrogen. These findings suggest that in mammals, oestrogen may still be a critical factor regulating ovarian somatic cell fate maintenance by preventing SOX9 nuclear translocation.

This fundamental role of oestrogen can explain its sex-determining ability in nonmammalian vertebrates. SOX9 remains cytoplasmic in the developing ovaries of nonmammalian vertebrates in the presence of oestrogen [5]. However, in the absence of oestrogen, SOX9 becomes nuclear and testicular development is initiated [41]. Our findings suggest that this may be a universal action of oestrogen in all vertebrates.

How oestrogen prevents SOX9 from entering the nucleus is still unknown. There are several known pathways that control SOX9 nuclear entry, such as altered importin-β1 or calmodulin binding, both of which facilitate effective transport of SOX9 across the nuclear pore complex [41, 42]. In addition, sumoylation and/or ubiquitination also affect the subcellular localization and transcriptional activity of SOX9 [49, 50]. Oestrogen is known to directly affect sumoylation and the subcellular localization of the nuclear receptor coactivator family member 3 protein in breast cancer cell lines [51]. Oestrogen may work in a similar way in the gonad by modulating sumoylation and/or ubiquitination of the SOX9 protein, trapping it in the cytoplasm.